Research And Application Of Point Of Interest Recommendation Algorithm Combining User Preference And Geographic Information

In the 21st century where the Internet developed at a high speed,there is a large number of data output every minute,which will inevitably cause the problem of information overload.The location-based social network(LBSN)takes a variety of application software as a carrier and selects information that meets user preferences from large amounts of data to recommend to users,providing users with a new experience that combines the virtual world and the physical world.The widespread use of smart mobile devices and the increasingly mature positioning technology make it easier to obtain the user’s real-time location,which makes location recommendation a focus which the experts and scholars research and use extensively.Personalized POI recommendations based on LBSN are performed by signing in users Modeling to help users analyze potential preferences and get recommendations for feedback to users.This article aims to improve the accuracy of recommendation results.The main tasks are as follows:(1)The research status of domestic and foreign scholars on LBSN is elaborated in detail.It mainly focuses on the three factors of geography,social and time.Basis on this,it makes clear the theoretical and practical significance of using LBSN for recommendation.(2)Conducting statistical analysis of the LBSN data(Gowalla data set),modeling users and POIs by analyzing user preferences and the characteristics of POI distance distribution,and finding out the regularity of sign-in behavior and the sign-in locations of users.(3)Perform K-Means clustering on the user’s check-in data to find the user’s preferred POI area,and use the power law distribution probability model to calculate the geographical influence on the distance between POIs.A hybrid recommendation that combines user preference and geographic information is proposed.(4)This article introduces adaptively adjusting the weights in the hybrid recommendation algorithm based on user activity and POI popularity,in order to solve the problem of poor recommendation results for inactive users and unpopular POIs.(5)A hybrid recommendation algorithm combining user preferences and geographic information is used as the core recommendation model to implement a mobile recommendation prototype system.